ClassifierXL: Example Classification of Geographical Data

As a result of neural classification of the table, ClassifierXL marks the rows of the table with a cluster number to which this row belongs. For example you have a table with geographical data that you need to classify.



Step 1: You click on ClassifierXL from the menu in MS Excel.


After the launch of the program, you will see the ClassifierXL dialog box.



Step 2: Using the mouse, choose the range of numerical data of the table that you want to classify (Source data), and indicate the cell number from which to start recording the cluster markers (Output).

Step 3. Set the following parameters:

- Number of Categories (clusters).
- Set autofilter. Setting this option allows you to review the table rows filtered by clusters.
- To One Column/Split Columns. This option allows to choose between two presentations: To One Column -- in one column named Clusters, for each row, you put the cluster number. Split Columns -- The columns are established by the number of clusters set in Number of Clusters option, for example, columns called Cluster 1, Cluster 2, Cluster 3, etc. In every column of each row, you put "1" if the row has fallen into this cluster and you put "0" if it hasn't.
- Learning rate: Number of epochs and Initial Neurons Weights set in recommended values. In most cases there is no need to change them.
- Set colors: Checking this option highlights each cluster with a different color.
- Calculate averages: Checking this box allows calculation of averages for each cluster. This is a very important characteristic of the cluster. The row of cluster averages, the differentiation of this row from the averages of other clusters and from the average values of the whole table, allows category judgments and additional knowledge mining from the table.
- Calculate minimum: Calculation of minimum values for each cluster.
- Calculate maximum: Calculation of maximum values for each cluster.
- Sort table by clusters: Checking this option allows you to group by clusters in the original table.


Step 4. Click on the Classify button and get the following results:



We have three clusters with the following specific weights.



You can see that 26% of countries fall in Cluster 1, 21% in Cluster 2, and 52% in Cluster 3. On the basis of Cluster Profile Chart analyses the following description could be made: Cluster 1 is a category of countries with big populations, population growth rate below the average, big land and water areas, many irrigated lands, and growth in Gross Domestic Product over the average. Cluster 2 is a category of countries with population above the average, population growth that is maximal, water and land areas are small but with irrigated land exceeding the average, and GDP growth somewhat below the average. Cluster 3 is a category of countries with populations 6 times lower than the average, standard population growth, water area considerably lower than average, area of irrigated land far below the average, and real GDP far below the average rate.